fredmiranda.com
Login

Moderated by: Fred Miranda
  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1              3              8       9       end
  

Archive 2016 · Canon 5D4 dynamic range analyzed from RAWs

  
 
garyvot
Offline
• • • • •
Upload & Sell: On
p.2 #1 · Canon 5D4 dynamic range analyzed from RAWs


level1photog wrote:
I just wish the price is more competitive. I'll be happy with Canon 6D II with the same sensor around ~$2000


I'm going to guess that the 6D II will come in at around 23Mp, with 80D-derived AF, and a 5fps framerate. Not bad, but still an entry level full frame body by contemporary standards. But hopefully they won't just repurpose the 5D III sensor...



Aug 30, 2016 at 02:55 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.2 #2 · Canon 5D4 dynamic range analyzed from RAWs


Just did it myself independently before seeing this, anyway cut and paste of my message on another thread (along with EDIT at the end after glancing at the DPR thread mentioned here for a second):

hmmmm just took a quick look at ISO 100 RAW file, after Canon this time mentioning DR so much for the first time I had hoped it would do even better than 1Dx2/80D, but it doesn't seem too. Much less a better 14.25 it doesn't even do the 13.75 I expected and seems likely to do more like 13.4 stops from what I can tell. A little disappointing TBH after all this time especially since the video quality at first tentative glances appears to be not quite as good as A7R II either. (A7R II at least 13.89 stops at ISO100, some places say better and more like 14-14.5; nikon 14.5-14.75 stops; might not all matter, except I fear the Canon might be a nice improvement and yet still just not quite enough to avoid having to combine more than one exposure since A7R II seems to only just barely handle the stuff I encounter, maybe the Canon can still sort of do it but there will be some bits of ugly showing a bit I think).

It is the single most cost efficient way to sort of get it all at once, but after all this time I thought they could at least tie some older stuff for stills and video quality. Not sure, but despite the negatives of a dual 5D3 + latest, greatest Sony combo, I may just stick with that (unless D820 utterly blow me away, not sure I can afford to swap to Nikon now though, $$ if you also need to swap over super-tele and had bought way back when they cost way, wayyyy less than they do today) as get better stills and video quality I think.

Not a bad camera, and certainly their best ever (well other than for 1DX2 ignoring the 20MP issue), and it is less expensive than pairing a 5D3 + adapter + Sony and you get the advantages of all in one system and get some extra DR even for action stuff and so on, but it does seem a little disappointing that after all these years they are still not caught up to Sony of a few years ago much less Nikon for low ISO stills max quality in high DR natural world scenarios (and again, simply keeping up doesn't even matter so much as being able to hit the critical minimum point to single shot handle stuff one often encounters in the real world, does it or not) and seemingly not quite caught up for video quality either (although the old 5D3 in a way is still the 1080P leader, so long as you can deal with RAW video, which does get to be a real drag after a while, that does deliver some fantastic quality and can even do unique things like wide gamut video). And it means at least another 3-4 years of waiting to get 14 stops. bleh. (unless they have a 5Ds mark II in another year or two and manage to get it to 6-7fps and top quality 4k) and they still insist on leaving out live 100%-200% zoom window during video/liveview which can be utterly critically important for basic usage.

Maybe the masked area is a touch noisier and the real sensor area will hit A7R II levels (which seems to me to be the bare minimum to just barely manage to handle lots of real world high DR forest and other such photography, maybe the new Canon pulls it off, but I'm a little afraid it might come up just a touch short to really quite pull off single frame shot, that tough stuff).

In many ways very good camera though, but it does seem a shame it didn't have just that final few extra bits though.

The supposedly calibrated LCD would be very cool and I'm sure DPAF could be nice at times, they fixed autoiso finally, fps isn't too bad, etc. etc. but the sensor still seems a bit blah and I could swear the video is just not showing the amazing, looking out a window clarity of the A7R II.

EDIT: apparently on DPR there is talk that DUalPixel RAW will somehow allow an extra stop of highlights though, haven't time to read it now, if actually true and it doesn't rob shadows (in which case it really isn't giving more highlights anyway) and there are no weird gotchas, that would be pretty cool news, at the cost of having to deal with 60MP files for a 30MP image you could then have the 13.4 stops turn into 14.4 stops?? IF, if that is actually true, then it suddenly does become much more interesting. 14.4 stops would be plenty good. You fire off typical shots in 30MP mode, use 60MP data size mode for the high DR non-action stuff (files too large for a good enough buffer for most sports; but still get a solid 13.4 for sports action). Hope this is really true.

Anyway, if it actually can pull of 14.4 in DualPixel RAW that would be pretty awesome and it would be a darn fine stills camera and a decent video one I guess, so not a bad package then.



Aug 30, 2016 at 03:54 PM
matejphoto
Offline
• • •
Upload & Sell: Off
p.2 #3 · Canon 5D4 dynamic range analyzed from RAWs


skibum5 wrote:
EDIT: apparently on DPR there is talk that DUalPixel RAW will somehow allow an extra stop of highlights though, haven't time to read it now, if actually true and it doesn't rob shadows (in which case it really isn't giving more highlights anyway) and there are no weird gotchas, that would be pretty cool news, at the cost of having to deal with 60MP files for a 30MP image you could then have the 13.4 stops turn into 14.4 stops?? IF, if that is actually true, then it suddenly does become much more interesting. 14.4 stops would be plenty good.
...Show more

This will be the big dilemma for the future owners (at least in the first couple of years before the benefits of DualPixel raw are hashed out):
Do you shoot everything (except fast faction) in DPRAW just in case if in the future they figure out a way to extract some extra "magic" things out of the files? I don't mind the size while shooting (7 image buffer is plenty and cards are big and cheap). My concern is processing 75MB files and long term storage. People say storage long term storage is cheap, but I already have over 3TB of pictures and all of it was done with sub 20Mpix cameras. DPRAW will burn though a terabyte fast.



Aug 30, 2016 at 04:23 PM
EB-1
Online
• • • • • • •
Upload & Sell: Off
p.2 #4 · Canon 5D4 dynamic range analyzed from RAWs


DP RAW is mainly for large apertures from what I read and results in noisier images. I certainly would not shoot everything on that mode. I already have the 5DsR for hi-res use. 10TB drives are reaching mainstream so storage should not be a huge concern.

EBH



Aug 30, 2016 at 05:11 PM
cgarcia
Offline
• •
Upload & Sell: Off
p.2 #5 · Canon 5D4 dynamic range analyzed from RAWs


I have found a 5D4 ISO 100 RAW with a bit more DR: about 13.58 EV (just +0.2 compared with my initial post). This is due because that RAW file had an average noise of 2.5 ADU instead of 2.88/3.07 I measured yesterday. Perhaps camera variation, or more likely temperature (Live View?) or similar factors.

I had no time today, but at least I already have downloaded 1DX2 and 5D4 RAWs from Imaging Resource to make a more complete comparison (to keep growing my already good collection from IR for another cameras). IR usually is faster than dpreview publishing the RAWs for the impatient (it seems that including this time, but I was no aware, due to an excess of impatience).



Aug 30, 2016 at 05:58 PM
cgarcia
Offline
• •
Upload & Sell: Off
p.2 #6 · Canon 5D4 dynamic range analyzed from RAWs


skibum5 wrote:
EDIT: apparently on DPR there is talk that DUalPixel RAW will somehow allow an extra stop of highlights though, haven't time to read it now, if actually true and it doesn't rob shadows (in which case it really isn't giving more highlights anyway) and there are no weird gotchas, that would be pretty cool news, at the cost of having to deal with 60MP files for a 30MP image you could then have the 13.4 stops turn into 14.4 stops?? IF, if that is actually true, then it suddenly does become much more interesting. 14.4 stops would be plenty good.
...Show more

Can't be true (outside technical stuff: if it were so easy to increase the DR by using smaller subpixels, all DSLRs long time ago would have been using phone-camera-sized pixels combined). The ADCs are more than capable to extract all the truly available DR, and nobody (except maybe God) can change noise into signal...



Aug 30, 2016 at 06:08 PM
cgarcia
Offline
• •
Upload & Sell: Off
p.2 #7 · Canon 5D4 dynamic range analyzed from RAWs


EB-1 wrote:
DP RAW is mainly for large apertures from what I read and results in noisier images. I certainly would not shoot everything on that mode. I already have the 5DsR for hi-res use. 10TB drives are reaching mainstream so storage should not be a huge concern.

EBH


And as dpreview has also shown, focus-shifted images are considerably less sharp in the critical focus point.

It also seems that the additional pixels can't provide extra resolution (as I ignorantly expected) since they both seem to "look at the same point".

The additional information stored in half of the data is already included in the other half (one subframe stores the combined pixels, and another subframe a single one). While I believe that image processing has still a long headroom in the future (demosaicing, sharpening...) I doubt of any revolutionary technology being able to go beyond from current Canon achievements with this new feature.

DPRAW offers only a tiny set of subtle features not totally for free. I think that many lovers of RAW archiving (as I in fact am) won't fall in love with it.



Aug 30, 2016 at 06:23 PM
garyvot
Offline
• • • • •
Upload & Sell: On
p.2 #8 · Canon 5D4 dynamic range analyzed from RAWs


cgarcia wrote:
And as dpreview has also shown, focus-shifted images are considerably less sharp in the critical focus point.

It also seems that the additional pixels can't provide extra resolution (as I ignorantly expected) since they both seem to "look at the same point".

The additional information stored in half of the data is already included in the other half (one subframe stores the combined pixels, and another subframe a single one). While I believe that image processing has still a long headroom in the future (demosaicing, sharpening...) I doubt of any revolutionary technology being able to go beyond from current Canon achievements with
...Show more

Canon states that the effect is more pronounced with longer focal lengths; DP Review inexplicably tested with a 35mm lens.

I don't expect miracles, but I'd say that writing off this feature based on their first test is premature.



Aug 30, 2016 at 06:58 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.2 #9 · Canon 5D4 dynamic range analyzed from RAWs


cgarcia wrote:
I have found a 5D4 ISO 100 RAW with a bit more DR: about 13.58 EV (just +0.2 compared with my initial post). This is due because that RAW file had an average noise of 2.5 ADU instead of 2.88/3.07 I measured yesterday. Perhaps camera variation, or more likely temperature (Live View?) or similar factors.

I had no time today, but at least I already have downloaded 1DX2 and 5D4 RAWs from Imaging Resource to make a more complete comparison (to keep growing my already good collection from IR for another cameras). IR usually is faster than dpreview publishing the RAWs
...Show more

One thing to watch out for is whether the shot was taken with a lenses set faster to f/2.8 or not, since they manipulate ISO in weird ways when fast lenses are used. If you test black frame from say and f/1.4 shot you won't get the proper results. (unless things are different on the 5D4, which I doubt)




Aug 30, 2016 at 07:13 PM
bhollis
Offline
• •
Upload & Sell: Off
p.2 #10 · Canon 5D4 dynamic range analyzed from RAWs


garyvot wrote:
Canon states that the effect is more pronounced with longer focal lengths; DP Review inexplicably tested with a 35mm lens.

I don't expect miracles, but I'd say that writing off this feature based on their first test is premature.


Actually, the two eyeball/eyelash tests were shot with an 85 at f/1.8 and 70-200 at 200mm and f/2.8. I'll wait for more definitive testing, but at this point, the usefulness of this feature appears minimal.



Aug 30, 2016 at 07:17 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.2 #11 · Canon 5D4 dynamic range analyzed from RAWs


cgarcia wrote:
Can't be true (outside technical stuff: if it were so easy to increase the DR by using smaller subpixels, all DSLRs long time ago would have been using phone-camera-sized pixels combined). The ADCs are more than capable to extract all the truly available DR, and nobody (except maybe God) can change noise into signal...


You need storage and CPU power to handle all those extras though and smaller process to make all those split or mini-sites so that explains that in part at least.

I still haven't had time to look it over, but doesn't that Illiah guy program RAW converters? I'd think he knows what he is talking about. It seemed like you could just treat the dual pixels as a single normal combined one for how you handle all of the image aside from the very top step beyond what the main channel provides, the main one seems to clip 1 stop before the extra channel so then you use the extra channel for the top bits (and who cares at that point if the light collecting area might be only 1/2, at the very top step, at ISO100, who cares about noise, you have more than enough light way up there).

So he seems to think you could simply set it so you just barely save the highlights as you normally would but then dial in an extra stop of exposure and then in post the extra stop that would be clipped can be saved from the extra channel while at the same time the extra stop exposure cleans up the shadow noise by a stop and thus you can turn say 13.4 stops (borderline marginal) into 14.4 stops (excellent and state of the art for ISO100 and, most importantly, enough to capture a lot of tricky scenes with a single exposure).

Again, I've been out and busy and haven't really looked over those posts and thought about it so maybe there is some bad flaw and it'll be stuck at 13.4, but if not, 14.4 is pretty nice (yeah you waste storage and store 60MP worth but not the end of the world, except for sports, but lots of sports cameras haven't had much more than the 13.4 stops yet anyway and you might still get enough buffer even with the '60MP' files for some wildlife work).

EDIT: I do wonder though if it might end up only grabbing light from half the incoming angles though for that top stop and might that not cause some issues?


Edited on Aug 30, 2016 at 08:05 PM · View previous versions



Aug 30, 2016 at 07:20 PM
Liquidstone
Offline
• • • •
Upload & Sell: Off
p.2 #12 · Canon 5D4 dynamic range analyzed from RAWs


cgarcia wrote:
The ADCs are more than capable to extract all the truly available DR, and nobody (except maybe God) can change noise into signal...


Wishing Canon would hire God soon, ahead of Nikon.



Aug 30, 2016 at 07:46 PM
bhollis
Offline
• •
Upload & Sell: Off
p.2 #13 · Canon 5D4 dynamic range analyzed from RAWs


cgarcia wrote:
Can't be true (outside technical stuff: if it were so easy to increase the DR by using smaller subpixels, all DSLRs long time ago would have been using phone-camera-sized pixels combined). The ADCs are more than capable to extract all the truly available DR, and nobody (except maybe God) can change noise into signal...


Here's the "technical stuff."

http://www.rawdigger.com/howtouse/Canon-dual-pixel-technology



Aug 30, 2016 at 08:06 PM
Pixel Perfect
Offline
• • • • • • •
Upload & Sell: Off
p.2 #14 · Canon 5D4 dynamic range analyzed from RAWs


EB-1 wrote:
DP RAW is mainly for large apertures from what I read and results in noisier images. I certainly would not shoot everything on that mode. I already have the 5DsR for hi-res use. 10TB drives are reaching mainstream so storage should not be a huge concern.

EBH


The discussion at dpreview seems to imply you would add +1EC and while the main frame will have blown highlights, the auxilliary frame - which is the capture from only half a pixel, will not be clipped and if the RAW converter can do it you could have effectively 1EV more DR and normal noise levels, from the main frame.

Hopefully we will see RAW software that can exploit the full potential of the Dual Pixel RAW file. DPP would be the initial hope but will they extend it's capabilities beyond the focus shift, flare reduction, bokeh shift?



Aug 30, 2016 at 09:00 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.2 #15 · Canon 5D4 dynamic range analyzed from RAWs


Pixel Perfect wrote:
The discussion at dpreview seems to imply you would add +1EC and while the main frame will have blown highlights, the auxilliary frame - which is the capture from only half a pixel, will not be clipped and if the RAW converter can do it you could have effectively 1EV more DR and normal noise levels, from the main frame.

Hopefully we will see RAW software that can exploit the full potential of the Dual Pixel RAW file. DPP would be the initial hope but will they extend it's capabilities beyond the focus shift, flare reduction, bokeh shift?


One might be able to write a custom converter that could take the highlights, pack them into the regular top values for the 5D4 and then slide the main values down 1 stop and then you could send the converted RAW to any old program and it would open it just like any old 5D4 regular RAW file only now the shadows would suddenly seem to be Nikon clean . And it would also have mid point and everything fitting as expected so no twisted profile color shifts or any other issues. Who knows when or if ACR or whatnot got around to doing it, so making a converter might get around such issues.

It seems a bit curious, if it really does work out, that Canon didn't do it to start, especially with all the talk about DR these days and even Canon being willing to mention it now. This could be a REAL HTP mode unlike on the old cameras, where it was sort of fake in that it didn't do anything you couldn't do by simply exposing 1 stop less. This one literally would truly save 1 stop of highlights in a way that nothing you could do in regular mode could ever replicate. Although since in the end sensors are linear and ultimately exposure it set to preserve the brightest part you want saved and all the drama is in the shadows it would really be a Shadow Noise Save Mode or an HDR mode where it gives you a full extra stop of DR for real. So why would they not have it already built in? (then again we are still dealing with pre-release stuff, maybe DPP ships with this mode, but you'd think they have mentioned it as a JPG option too in camera firmware).




Edited on Aug 30, 2016 at 10:35 PM · View previous versions



Aug 30, 2016 at 10:31 PM
EB-1
Online
• • • • • • •
Upload & Sell: Off
p.2 #16 · Canon 5D4 dynamic range analyzed from RAWs


Pixel Perfect wrote:
The discussion at dpreview seems to imply you would add +1EC and while the main frame will have blown highlights, the auxilliary frame - which is the capture from only half a pixel, will not be clipped and if the RAW converter can do it you could have effectively 1EV more DR and normal noise levels, from the main frame.

Hopefully we will see RAW software that can exploit the full potential of the Dual Pixel RAW file. DPP would be the initial hope but will they extend it's capabilities beyond the focus shift, flare reduction, bokeh shift?


That's all far too complicated for a mere 30MP camera. I mainly wanted a body to use at ISO 1600-3200 with a better framing rate than the 5DsR and higher resolution than the 1DX/1DX II. The 5DsR has excellent IQ at the low ISOs. No DPRAW will overcome the 50MP and no AA filter for me.

EBH



Aug 30, 2016 at 11:38 PM
garyvot
Offline
• • • • •
Upload & Sell: On
p.2 #17 · Canon 5D4 dynamic range analyzed from RAWs




bhollis wrote:
Actually, the two eyeball/eyelash tests were shot with an 85 at f/1.8 and 70-200 at 200mm and f/2.8. I'll wait for more definitive testing, but at this point, the usefulness of this feature appears minimal.

Well if that's the case I stand corrected and I would tend to agree.



Aug 31, 2016 at 12:00 AM
Pixel Perfect
Offline
• • • • • • •
Upload & Sell: Off
p.2 #18 · Canon 5D4 dynamic range analyzed from RAWs


I wonder if Magic Lantern could do a custom fw that could combine the highlights from the auxilliary frame with the rest of the data from the main frame and output a standard size cr2 file, thast would be sweet.

It looks like the dual pixel raw may be of use outside what Canon intended, since the effects it creates are very subtle. I'd only see a use for flare reduction in the real world, the refocus effect is equal to about +/-2 for MFA in camera, so very subtle and bokeh shift is weird, maybe you could use it to shift a very distracting oof highlight but the shift amout is small. However, being able to get 14.4EV of DR would be stellar.



Aug 31, 2016 at 12:47 AM
EvilZardoz
Offline
• •
Upload & Sell: Off
p.2 #19 · Canon 5D4 dynamic range analyzed from RAWs


Exciting times! I do a lot of available light work, so having the ability to control my dynamic range in post is something that is huge for me.

Big concern re: CPU/GPU performance. I like to use my MacBook Pro for image editing so I can take it on the road and have everything with me (20 minutes to spare? That's 20 minutes of cranking out more images in post). Apple haven't shipped a significant performance upgrade for the last three years on the MacBook Pros and the same deal with the Mac Pro. Unless Apple drop something amazing with their next refresh, I feel I'll be very frustrated with DPraw images. Experience has told me to keep every raw image though.

Storage is cheap, though - so that doesn't concern me.



Aug 31, 2016 at 10:06 AM
Hathaway
Offline
• • •
Upload & Sell: Off
p.2 #20 · Canon 5D4 dynamic range analyzed from RAWs


charlesk wrote:
Thanks. I wish I had a better subjective feel for what actual impact that will have in the "real world".


For me I am hoping for the following real world improvement over the 5DSR. Increased ability to shoot at higher ISOs without as much of a noise penalty. Above ISO 1600 I start to get a lot of noise with the 5DSR that is difficult to deal with or causes the images to be unusable. This is primarily for wildlife. 5DSR is still a beast for landscape even though more DR would help in that situation you can always bracket for more DR.

In addition, it will help with harsh light situations for wildlife. These conditions create lots of shadows on birds, etc. that have to be lifted and start to introduce noise. I will be happy if there is more DR and less of a noise penalty for my shots.

I am really interested in comparisons between the 5DMIV and the 1DXII. I know I am getting extra MP in exchange for lower FPS, but I am wondering how they perform for ISO, DR and noise.



Aug 31, 2016 at 10:15 AM
1              3              8       9       end




FM Forums | Canon Forum | Join Upload & Sell

1              3              8       9       end
    
 

Welcome back
Log in to your account